Fix hebrew character highlight issue in a special document

core/src/fpdftext/fpdf_text_int.cpp

Index: core/src/fpdftext/fpdf_text_int.cpp
diff --git a/core/src/fpdftext/fpdf_text_int.cpp b/core/src/fpdftext/fpdf_text_int.cpp
index de4c703a042635d2c195443ac9c1e3398e05dbb7..5b174f0187bf523fe77675d86854f0856b4e25ae 100644
--- a/core/src/fpdftext/fpdf_text_int.cpp
+++ b/core/src/fpdftext/fpdf_text_int.cpp
@@ -1682,6 +1682,53 @@ void CPDF_TextPage::ProcessTextObject(PDFTEXT_Obj Obj)
             baseSpace = 0.0;
         }
     }
+
+    FX_BOOL bIsBidiAndMirrosInverse = FALSE;

[jbreiden, 2014/08/18 19:10:50]

What does MirrosInverse mean?

+    IFX_BidiChar* BidiChar = IFX_BidiChar::Create();
+    FX_INT32 nR2L = 0;
+    FX_INT32 nL2R = 0;
+    FX_INT32 start = 0, count = 0;
+    CPDF_TextObjectItem item;
+    for (FX_INT32 i = 0; i < nItems; i++) {
+        pTextObj->GetItemInfo(i, &item);
+        if (item.m_CharCode == (FX_DWORD)-1) {
+            continue;
+        }
+        CFX_WideString wstrItem = pFont->UnicodeFromCharCode(item.m_CharCode);
+        FX_WCHAR wChar = wstrItem.GetAt(0);
+        if ((wstrItem.IsEmpty() || wChar == 0) && item.m_CharCode) {

[jbreiden, 2014/08/18 19:10:49]

Same question. Is the character to Unicode mapping from the PDF file used?

[Bo Xu, 2014/08/19 03:31:49]

Yes, in CFX_WideString CPDF_Font::UnicodeFromCharCode(FX_DWORD charcode) const,
it does this :

    if (!m_bToUnicodeLoaded) {
        ((CPDF_Font*)this)->LoadUnicodeMap();
    }

to load the UnicodeMap if present.

[jbreiden, 2014/08/22 16:58:44]

Okay.  If I understand correctly:

The PDF files I provided have a UnicodeMap with full coverage. Therefore
UnicodeFromCharCode(item.m_CharCode) is successful, and line 1700 never
executes.

Looks good to me.

+            wChar = (FX_WCHAR)item.m_CharCode;
+        }
+        if (!wChar) {
+            continue;
+        }
+        if (BidiChar && BidiChar->AppendChar(wChar)) {
+            FX_INT32 ret = BidiChar->GetBidiInfo(start, count);
+            if (ret == 2) {
+                nR2L++;
+            }
+            else if (ret == 1) {
+                nL2R++;
+            }
+        }
+    }
+    if (BidiChar && BidiChar->EndChar()) {
+        FX_INT32 ret = BidiChar->GetBidiInfo(start, count);
+        if (ret == 2) {
+            nR2L++;
+        }
+        else if (ret == 1) {
+            nL2R++;
+        }
+    }
+    FX_BOOL bR2L = FALSE;
+    if (nR2L > 0 && nR2L >= nL2R) {
+        bR2L = TRUE;
+    }
+    bIsBidiAndMirrosInverse = bR2L && (matrix.a * matrix.d - matrix.b * matrix.c) < 0;
+    FX_INT32 iBufStartAppend = m_TempTextBuf.GetLength();
+    FX_INT32 iCharListStartAppend = m_TempCharList.GetSize();
+
     for (int i = 0; i < nItems; i++) {
         CPDF_TextObjectItem item;
         PAGECHAR_INFO charinfo;
@@ -1828,6 +1875,30 @@ void CPDF_TextPage::ProcessTextObject(PDFTEXT_Obj Obj)
             }
         }
     }
+    if (bIsBidiAndMirrosInverse) {

[jbreiden, 2014/08/18 19:10:50]

Ok, I see what is going on here. The code is reversing some
text when necessary because of BiDi. This looks like a 
heuristic.

Please, please, please let me know if there is a mistake 
with the sample PDF file. I have 100% control over the PDF 
generation and can fix them if there is a mistake. This is 
better than PDFium being "smart" and using heuristics to 
work around bad PDF files.

[Bo Xu, 2014/08/19 03:31:49]

I wonder do you generate this particular pdf file? This is indeed a very special
file, it has an image object that shows the correct text, but the actual text
objects are stored in a mirror fashion which makes the text selection goes
wrong.

[jbreiden, 2014/08/22 16:58:44]

The PDF file is generated by Tesseract, which is an open source OCR program. I
am the engineer responsible for this feature and have the ability to change it.

https://code.google.com/p/tesseract-ocr/

I have no further comments, except to encourage the best possible solution. In
the future I will be working on PDF output for several other programs, so it is
good to work trouble now.

+        FX_INT32 i, j;
+        i = iCharListStartAppend;
+        j = m_TempCharList.GetSize() - 1;
+        PAGECHAR_INFO tempCharInfo;
+        FX_INT32 tempIndex = 0;
+        for (; i < j; i++, j--) {
+            tempCharInfo = m_TempCharList[i];
+            m_TempCharList[i] = m_TempCharList[j];
+            m_TempCharList[j] = tempCharInfo;
+            tempIndex = m_TempCharList[i].m_Index;
+            m_TempCharList[i].m_Index = m_TempCharList[j].m_Index;
+            m_TempCharList[j].m_Index = tempIndex;
+        }
+        FX_WCHAR * pTempBuffer = m_TempTextBuf.GetBuffer();
+        i = iBufStartAppend;
+        j = m_TempTextBuf.GetLength() - 1;
+        FX_WCHAR wTemp;
+        for (; i < j; i++, j--) {
+            wTemp = pTempBuffer[i];
+            pTempBuffer[i] = pTempBuffer[j];
+            pTempBuffer[j] = wTemp;
+        }
+    }
 }
 FX_INT32 CPDF_TextPage::GetTextObjectWritingMode(const CPDF_TextObject* pTextObj)
 {